Facebook announced CompilerGym open source this week as part of its efforts to improve compiler performance by leveraging machine learning to tackle optimization work.
CompilerGym is built by the Facebook AI team atop the OpenAI Gym and ultimately strives to help improve the performance of the code compiler. They explained in this week’s announcement, “CompilerGym bundles important compiler optimization problems together and makes them look like reinforcement learning problems. The compiler optimization problems that we include are large scale. For example, for one, the search space is 104461, considerably larger than that of the board game Go. On the other hand, the search space is infinite. Progress on problems of this magnitude is only possible for the first time because of the very recent advances in reinforcement learning. CompilerGym allows anyone with an ML or compiler background to jump in and start troubleshooting, all without the months of tedious setup time that would normally be required. And that’s because we did it for you!“
The ad added: “Our goal is to be a catalyst for using ML to make compilers faster, which is important because poorly optimized programs are slow and consume too much computing resources as well as too much power, limiting device applications. energy efficient edge devices and downsizing data centers. ecological.“
Those who wish to learn more about CompilerGym can do so via ai.facebook.com.